-
FINANCIALREPORTINGSECTION
26 IFRS 17 Variable Fee ApproachBy Tze Ping Chng, Steve Cheung
and Anson Yu
32 Briefly Noted: Goings on in Reserving and Modeling for
A&H Waiver of PremiumsBy Xianmei Tang, Anthony Muturi, Shanpi
Yu and Isaac Larbi
33 Alert on New Valuation Rate Methodology for Payout Annuities
and Similar ContractsBy Paul Hance and Heather Gordon
34 Financial Reporting Research UpdateBy David Armstrong and
Ronora Stryker
3 Chairperson’s CornerBy Bob Leach
4 FASB Long-Duration Contracts RedeliberationsBy Leonard
Reback
10 GAAP Targeted Improvements—Unlocking PersistencyBy Steve
Malerich
14 Indexed Variable Annuities: The Next Product Frontier for the
U.S. Annuity MarketBy Simpa Baiye, Robert Humphreys and David
Knipe
18 Asset Modeling Challenges for VM-20 ProjectionsBy Ben
Slutsker, Jason Kehrberg and Reanna Nicholsen
23 Leveraging X-factor Testing Techniques in Developing
Mortality Assumptions for VM-20By Je rey R. Lortie and Ying
Zhao
FASB Long-Duration Contracts RedeliberationsBy Leonard
Reback
Page 4
The FinancialReporter
ISSUE 112 • MARCH 2018
-
2 | MARCH 2018 THE FINANCIAL REPORTER
The Financial Reporter
2018 SECTION LEADERSHIP
OfficersBob Leach, FSA, MAAA, Chairperson Simpa Baiye, FSA,
MAAA, Vice Chairperson Enzinma Miller, FSA, MAAA, SecretaryDavid
Ruiz, FSA, FIA, MAAA, Treasurer
Council Members David Armstrong, FSA, MAAALance Berthiaume, FSA,
MAAAKatie Cantor, FSA, MAAASteve Finn, FSA, MAAAAshwini Vaidya,
FSA, MAAA
Newsletter EditorMichael Fruchter, FSA, MAAA
[email protected]
Associate EditorsAisling Metcalfe, FSA, FIA,
[email protected]
Marc Whinston, FSA, [email protected]
Program Committee CoordinatorsDavid Armstrong, FSA, MAAALance
Berthiaume, FSA, MAAAKatie Cantor, FSA, MAAA 2018 Valuation Actuary
Symposium Coordinator
Simpa Baiye, FSA, MAAAEnzinma Miller, FSA, MAAADavid Ruiz, FSA,
FIA, MAAA2018 Life & Annuity Symposium Coordinators
Katie Cantor, FSA, MAAASteve Finn, FSA, MAAA Ashwini Vaidya,
FSA, MAAA2018 SOA Annual Meeting & Exhibit Coordinators
SOA StaffJim Miles, Staff [email protected]
Jessica Boyke, Section Specialist [email protected]
Julia Anderson Bauer, Publications Manager
[email protected]
Sam Phillips, Staff [email protected]
Julissa Sweeney, Graphic Designer [email protected]
Published quarterly by the Financial Reporting Section of
the
Society of Actuaries.
475 N. Martingale Road, Suite 600Schaumburg, Ill 60173- 2226
Phone: 847. 706. 3500 Fax: 847. 706. 3599www.soa.org
This newsletter is free to section members. Current issues are
available on
the SOA website (www.soa.org).
To join the section, SOA members and non- members can locate a
membership form on the Financial Reporting Section
Web page at http://www.soa.org/sections/financial-reporting/
financial-reporting-landing/.
This publication is provided for informational and
educational
purposes only. Neither the Society of Actuaries nor the
respective authors’
employers make any endorsement, representation or guarantee
with
regard to any content, and disclaim any liability in connection
with the
use or misuse of any information provided herein. This
publication
should not be construed as professional or financial advice.
Statements of
fact and opinions expressed herein are those of the individual
authors
and are not necessarily those of the Society of Actuaries or
the
respective authors’ employers.
Copyright © 2018 Society of Actuaries.All rights reserved.
Publication Schedule Publication Month: September
Articles Due: June 18
Issue Number 112 • March 2018
-
MARCH 2018 THE FINANCIAL REPORTER | 3
Chairperson’s CornerBy Bob Leach
Implementation of the NAIC Valuation Manual (VM) for 2017
reporting reminds us of the many things we must do in order to be
good financial reporting actuaries. Some may view these activities
as merely a distraction. Nothing could be further from the truth.
It is in the attending to these details that the financial
reporting actuary engenders confidence among associates, regulators
and other users of our statutory financial reports, and gains
distinction as a true professional. The new requirements will also
result in improved governance and understanding of our work and
increased appreciation from others for the responsibilities of the
financial reporting actuary. For example, consider the
following:
• VM-05 (the NAIC Model Standard Valuation Law) requires that
the assumptions and methods used in principle-based reserve (PBR)
valuation are consistent with the company’s risk assessment
process. It also requires a certification of PBR control
effectiveness to the company’s board of direc-tors and its
domiciliary regulator. This solidifies the link that should exist
among three important processes: valuation, risk management and
controls. It also helps to increase the board’s understanding and
engagement in our work.
• VM-30 (Actuarial Opinion and Memorandum Require-ments)
requires the Actuarial Opinion to include a Table of Key
Indicators, making it easier for the regulator to deter-mine
whether the appointed actuary’s opinion regarding reserve adequacy
is unqualified, and alerting the regulator to the use of wording in
the opinion that is other than that pre-scribed in VM-30. The VM
creates additional requirements for regulatory actuaries too, so it
makes sense to minimize obstacles they might otherwise face in
understanding a com-pany’s Actuarial Opinion.
• VM-G (Corporate Governance Guidance for Princi-ple-Based
Reserves) spells out specific responsibilities with respect to PBR
valuation for the company’s board of directors, senior management
and the qualified actuary (or
actuaries) providing certification of PBR reserves. Clarity with
respect to the roles played by each party helps all of those
involved in the PBR process to ensure that they have appropriately
fulfilled their responsibilities.
To be sure, these and other aspects of the VM have created a lot
of new requirements. This year-end was busier than last, and for
companies which have elected to defer implementation of VM-20
(Requirements for Principle-Based Reserves for Life Products) to a
future year, there is even more work ahead. Also remaining ahead
for all of us is the implementation of VM-50 (Experience Reporting
Requirements).
Paying attention to these details can provide a learning
experi-ence, and it helps to approach the new requirements with
this mindset. I have often found that documentation of a proce-dure
or execution of a control can uncover opportunities to enhance
actuarial modeling. The documentation, governance and control
requirements laid out in the VM are not a dis-traction—rather, they
are a means to improve and validate the quality of our actuarial
work. And quality work is the hallmark of professionalism!
Bob Leach, FSA, MAAA, is a vice president at Fidelity
Investments Life Insurance Company. He can be reached at
[email protected].
-
4 | MARCH 2018 THE FINANCIAL REPORTER
FASB Long-Duration ContractsRedeliberationsBy Leonard Reback
The Financial Accounting Standards Board (FASB) was busy in the
second half of 2017 redeliberating decisions made under their
long-duration contracts accounting project for insurance companies.
FASB promulgates Gen-erally Accepted Accounting Principles (GAAP)
for general reporting purposes in the United States. FASB has been
working on a project to update and improve accounting for insurance
contracts for almost 10 years now. In 2015 it issued new guidance
for short-duration contracts, requiring several additional
disclosures. It is now approaching the finish line on its
long-duration contracts project, and is expected to issue a new
standard updating both disclosure and measurement of insurance
contracts in 2018.
FASB had issued an exposure draft (ED) of its tentative
deci-sions on long-duration contracts in September 2016. After
receiving 39 formal comment letters responding to the ED,
performing outreach with financial statement users and holding a
roundtable discussion in April, FASB began redeliberating its ED
proposals in August. Two more meetings followed in Octo-ber and
November. As a result of redeliberations, FASB made several key
changes to its previous decisions. The basic scope of the proposed
changes remains similar, however. As of December 2017, it appears
that all major decisions have been made except for determining the
effective date of the new standard, although no decisions are final
until the standard is issued. The major changes that had been
decided through December are discussed in this article.
TRADITIONAL NON-PARTICIPATING INSURANCE CONTRACT RESERVESUnder
current US GAAP, traditional non-participating insur-ance contracts
(FAS 60 and FAS 97 limited pay) hold net premium reserves based on
assumptions that are locked in when the contract is issued unless a
premium deficiency emerges. The assumptions, including the expected
investment return that is used as the discount rate, include a
provision for adverse devi-ation (PAD), which incorporates some
conservatism into the
reserve. A premium deficiency test is required periodically to
ensure that the reported reserve is not inadequate.
Under the ED proposals, cash flow assumptions would be reviewed
for possible updates at least annually. When assump-tions are
updated, the net premium ratio (and any deferred profit liability
for limited pay contracts) would be updated retrospec-tively. That
is, the net premium ratio would be reset assuming all actual
historical experience, as well as the new assumptions, had been
known since the contract was issued. This process is similar to
current US GAAP accounting for deferred acquisition costs (DAC) on
universal life contracts. The net premium ratio would be subject to
a cap of 100 percent. One subtle change to the cash flow
assumptions is that the ED eliminated most maintenance expenses
from the reserve calculations, retaining only such non-level
expenses as claim costs.
To the extent that the net premium ratio changes, that would
offset part of the impact of the present value of future cash flows
on the reserve. But the change in present value of future cash
flows that would not be offset by unlocking the net premium ratio
would impact the reserve immediately, with a correspond-ing impact
to net income. Because the assumptions would be updated, provisions
for adverse deviation were eliminated. And because the net premium
ratio would be subject to a 100 percent cap, premium deficiency
testing was eliminated.
In the ED, FASB proposed to treat the discount rate differently.
FASB proposed using a more market-based objective discount rate
than the expected investment (i.e., “book”) yield, feeling that it
was not appropriate for a non-participating liability value to be
impacted by expected asset performance. FASB proposed discounting
the liability using a “high-quality fixed-income yield,” generally
interpreted to mean a AA-quality bond yield. The discount rate
would be updated each reporting period. The impact of changing the
discount rate would be reported in other comprehensive income (OCI)
without impacting the net premium ratio. Reporting the change in
discount rates through OCI was deemed to avoid accounting
mismatches with the assets insurers hold to back such liabilities,
which typically report changes in fair value due to changes in
interest rates through OCI.
Many companies and industry groups objected to the ED pro-posal
to retrospectively unlock the net premium ratio. They felt that
this would be costly to implement and would result in unnecessary
net income volatility. Many comment letters proposed using a
prospective unlocking approach instead, sim-ilar to the ED
proposals for DAC. Many comment letters also objected to using a AA
discount rate, feeling that such a rate was overly conservative and
did not provide an adequate illiquidity premium.
-
MARCH 2018 THE FINANCIAL REPORTER | 5
In response, the board made a number of changes during
redeliberations. The board felt that a retrospective unlocking
approach provided the most relevant measure for a liability that
represents a future cash flow. As a result, FASB retained
retrospective unlocking for the net premium ratio. But it did make
a number of changes to make the process somewhat less operationally
burdensome.
FASB recognized that a significant portion of the cost of
retro-spective unlocking for universal life DAC relates to
allocating items such as expenses and investment income to
contracts. The proposed calculation of non-participating contract
reserves would already not require an allocation of investment
income and only a limited amount of expense would be permitted in
the reserve calculation. So FASB decided to eliminate the
require-ment to unlock the remaining expense assumptions, leaving a
company an option on whether or not to do so. FASB also rec-ognized
that much of the cost of retrospective unlocking relates to truing
up actual experience, as opposed to just updating assumptions. So
FASB decided to eliminate the requirement for companies to true-up
actual experience each reporting period, permitting companies to
choose to only true-up actual experi-ence once a year at the same
time as assumption updates. FASB also simplified the transition
requirements for these contracts, as will be discussed in the
“Transition” section of this article.
With respect to discount rates, FASB retained the requirement to
update the discount rate each reporting period and report the
impact of the change through OCI. But FASB agreed with the comment
letters stating that a AA discount rate was overly conservative and
decided to require an “upper-medium grade fixed income yield,”
generally interpreted as a single-A quality discount rate.
TRADITIONAL PARTICIPATING CONTRACT RESERVESThe ED proposed that
participating contract (FAS 120) reserves (including those for
closed blocks) be calculated in a manner similar to the proposed
approach for non-participating reserves. Many comment letters
objected on the basis that the proposed model was not suited to the
unique features of participating contracts. For example, the
proposed model would ignore the link between the investment returns
on assets backing the lia-bility and the dividend cash flows of the
liability. In response to these comments FASB decided to exclude
FAS 120 contract reserves from the scope of the targeted
improvement project. Thus, FAS 120 reserves would continue to be
calculated as they are currently, including the need for a premium
deficiency test (without the inclusion of DAC). There would likely
be some minor changes to accounting for these contracts to conform
to other aspects of the targeted improvements, such as simplified
DAC amortization. For example, currently terminal dividend
liabilities are accrued over estimated gross margins (EGMs). With
EGMs being eliminated from the DAC model, terminal
dividend liabilities would likely be accrued using the new basis
for amortizing DAC.
UNIVERSAL LIFE CONTRACT RESERVESThe ED proposed significant
changes to the calculation of SOP 03-1 reserves for additional
death and annuitization benefits on universal life contracts. As
with the participating contract reserve proposals, comment letters
convinced FASB that the proposal would not work as intended. As a
result, FASB decided to largely retain the existing approach to
calculating SOP 03-1 reserves. There would likely be some minor
conforming changes. For example, the discount rate to use for
discounting payout annuity benefits back to the anticipated
annuitization date would be the single-A “upper-medium grade fixed
income yield,” consistent with the discount rate for
non-participating reserves.
Since the universal life contract valuation model would remain
essentially unchanged, the premium deficiency test would con-tinue
to be required, albeit excluding DAC.
DAC AND SIMILAR ITEMSUnder current US GAAP there are multiple
approaches to amortize DAC (and similar items such as deferred
sales induce-ments and unearned revenue). Depending on which
accounting model the underlying contracts fall into, DAC is
amortized in proportion to premiums, estimated gross profits,
estimated gross margins or in some cases in proportion to some
other contract element, such as death benefits. Some DAC models use
locked-in assumptions, others use retrospective unlocking. Some
investment contracts use an effective yield approach to amortize
DAC.
In the ED, FASB proposed to conform almost all DAC approaches,
the exception being retaining the effective yield approach for
certain investment contracts. FASB proposed to amortize DAC for all
other contracts in proportion to amount of insurance, or if amount
of insurance cannot be projected then on a straight line basis.
Assumptions would be unlocked prospectively; that is, when future
assumptions of terminations change, the future DAC amortization
schedule would “pivot” to reflect the revised assumptions, but the
current balance would not change. Interest would no longer be
accrued on DAC or similar items. The amortization ratio would not
be permitted to anticipate future renewal expenses or front-end
fees. Rather, the
FASB decided to largely retain the existing approach to
calculating SOP 03-1 reserves.
-
6 | MARCH 2018 THE FINANCIAL REPORTER
Redeliberations
amortization ratio would be updated as the new expenses were
incurred, so the amortization ratio could increase over time even
if experience materialized exactly as expected. DAC would no longer
be tested for impairment.
In its redeliberations FASB retained most of their ED decisions.
However, in response to comments that amount of insurance in force
is not necessarily an appropriate amortization approach for all
contract types, FASB agreed to be less restrictive. As a result,
DAC and similar items would be amortized in constant proportion to
some contract element (or straight line), but the contract element
would not necessarily need to be the amount of insurance in force.
DAC and similar items would still not accrue interest or be subject
to impairment testing, and future renewal costs or front-end fees
would still not be anticipated in the amortization ratio. Some
actuaries remain concerned about the latter issue with respect to
front-end loads in situations where the front-end fees are charged
over an extended period, potentially resulting in an amortization
ratio that increases sig-nificantly over time.
Some comment letters noted the irony that FASB was elimi-nating
retrospective unlocking for DAC, partially in response to concerns
from companies over cost and from users over incomprehensibility.
On the other hand, FASB was introducing retrospective unlocking for
non-traditional contract reserves. FASB seems to believe that a
retrospective unlocking approach is appropriate for changes in
future cash flows, and that the resulting volatility is meaningful
as an improved measurement of the present value of future cash
flows. However, FASB seemed to agree that retrospective unlocking
of DAC, which represents a cash flow that has occurred in the past,
is not particularly meaningful. In particular, FASB seemed
concerned about the practice of amortizing DAC and then potentially
reestablishing it through an unlocking event.
MARKET RISK BENEFITSThe ED introduced a new concept of a market
risk benefit (MRB). This concept would apply to guarantees on
certain variable contracts that expose the insurer to other than
nom-inal capital market risk. In particular, guaranteed minimum
death, income, withdrawal and accumulation benefits on qual-ifying
variable contracts would be MRBs. Also, many variable life no-lapse
guarantees would be MRBs. If a guarantee was
considered an MRB, the benefit would be reported at fair value.
Changes in fair value would be reported in net income, except for
changes in fair value resulting from changes in own credit which
would be reported in OCI. This accounting would apply regardless of
whether the guarantee is considered an embedded derivative under
current US GAAP.
In its redeliberations FASB expanded the scope of MRBs to go
beyond just variable contracts. The revised scope seems to
encompass guaranteed minimum death, income, withdrawal and
accumulation benefits on both variable and indexed con-tracts. The
equity indexing feature which is currently typically reported as an
embedded derivative on EIA and EIUL contracts also appears to be
within the revised MRB scope. However, FASB focused the revised
scope on account balance guarantees, which may scope out variable
life no-lapse guarantees. The revised basic definition of an MRB
(excluding some explanatory language) as disclosed at the November
2017 FASB meeting is as follows:
In its redeliberations FASB expanded the scope of MRBs to go
beyond just variable contracts.
-
MARCH 2018 THE FINANCIAL REPORTER | 7
“A market risk benefit shall be recognized for a contract
feature that exposes the insurance entity to other-than-nominal
capital market risk that arises from either of the following:
a. a contract feature that protects the account balance (or
simi-lar amount) from adverse capital market performance or
b. a contract feature that causes variability in the account
balance (or similar amount) in response to capital market
volatility.”
It is not entirely clear which other insurance contract features
would be scoped into this definition. It is possible that the
definition may be refined and further clarified when the new
accounting standard gets drafted in order to ensure that FASB
scopes in the features it intends without scoping in other
features.
DISCLOSURESThe ED proposed requiring many new footnote
disclosures. In response to comment letter feedback FASB decided to
eliminate a few of the more onerous requirements. But many new
foot-note disclosures would be added.
Most notably, roll-forwards would be required for all reserve
and DAC balances. Information about assumptions and changes in
assumptions would be required, as well as information about the
impact of assumption changes on the reserve balances. For
traditional non-participating contracts, information would be
required about the gross premiums, net premiums and benefits,
including their undiscounted amounts. For universal life con-tracts
a table would be required showing guaranteed and current credited
rates. For market risk benefits, information would be required
about benefits whose fair value is an asset versus a liability.
Disclosures would be required for non-participating traditional
contracts whose net premium ratio gets capped at 100 percent and
for other contracts that fail a premium defi-ciency test. And there
would be other requirements as well. There may be some changes to
the requirements as FASB gets feedback from users on their recent
decisions, particularly on ED requirements that were
eliminated.
TRANSITIONFASB made some minor and some major changes to the
transi-tion requirements from the ED. The most significant changes
were to transition for non-participating reserves. Under the ED,
non-participating reserves would have been required to use a
retrospective transition. That is, the reserve would have had to be
calculated since the contract was issued as if the new guid-ance
had been in effect all along. Only if it was “impracticable” to
determine or estimate the historical information necessary could a
prospective transition be used. Under a prospective tran-sition,
the existing GAAP balance on the transition date would
carry over (after removing any amounts that had been reported
through OCI) and the net premium ratio would be calibrated to the
reserve balance on the transition date. When assumptions would be
updated in the future, the retrospective unlocking of the net
premium ratio would go back to the transition date, not the
original issue date.
FASB decided to change the ED proposal to instead require a
prospective transition for all non-participating contract reserves.
FASB is allowing an option to use a retrospective transition, but
with several strings attached:
a. A company must be able to use actual historical data in order
to apply retrospective transition; the historical information may
not be estimated, and
b. a company must retrospectively transition all contracts
issued in a given year or later.
For example, if a company had actual historical information for
all contracts issued from 2014 and later, it would be permitted to
use retrospective transition for all contracts issued in 2014 or
later. It could choose a later issue date for which to apply
retrospective transition, but not an earlier date. It could not
retrospectively transition contracts issued in 2014 but
prospec-tively transition contracts issued in 2016. Any contracts
older than 2014 (or whatever year was chosen for retrospective
transi-tion) would have to be transitioned prospectively.
For DAC and similar balances the ED had proposed a pro-spective
transition. FASB mostly retained this decision, but conformed the
decision to the non-participating contracts deci-sion. So, if a
company decided to retrospectively transition all non-participating
contracts issued in 2014 and later, it would also need to
retrospectively transition all DAC for all contracts (including
other types of contracts) issued in 2014 and later. If the company
did not have the actual data to retrospectively tran-sition all DAC
on 2014 issues, it would also not be permitted to retrospectively
transition non-participating contracts issued in 2014.
FASB also made a small but possibly significant change to the
transition requirements for market risk benefits. The ED had
required a retrospective transition. That is, the attributed fee
associated with the market risk benefit would need to be
cali-brated to conditions as of the issue date of the contract.
Many comment letters argued that this was an onerous requirement
and also expressed concern that this could dramatically increase
the reserve for these benefits upon transition, thus materially
reducing GAAP equity. Comment letters also argued that it was
unrealistic to require an actuary to estimate an attributed fee for
a contract issued in, say 2006, and calibrate stochastic scenarios
to do so pretending to be unaware of future dramatic events
that
-
8 | MARCH 2018 THE FINANCIAL REPORTER
Redeliberations
Leonard J. Reback, FSA, MAAA, is vice president and actuary at
Metropolitan Life Insurance Company in Bridgewater, N.J. He can be
reached at [email protected].
had actually subsequently occurred, such as the significant
stock market declines in 2008/2009 and negative interest rates.
FASB gave some relief to the latter issue by still requiring a
retrospective transition, but permitting the actuary to use
“hind-sight” when calibrating the necessary scenarios. It is not
entirely clear that this resolves all the practical issues, and
this may not give much if any relief from the possible hit to GAAP
equity upon transition.
CONCLUSIONBig changes are coming to GAAP accounting for
long-duration contracts for insurance companies. FASB seems
determined to conclude this project as quickly as possible, and so
a final stan-dard is expected in 2018, possibly in early 2018.
Although we do not yet know when the new standard would be
effective, we do
know we would need to change our valuation models for several
reserve categories. Valuation of non-participating traditional
contract reserves is likely to become much more complicated. Many
benefits on variable and indexed contracts that are not fair valued
today would need to be fair valued in the future. DAC amortization
would become simpler but there would still be one-time changes
needed to the amortization models. And many more disclosures would
be required.
-
10 | MARCH 2018 THE FINANCIAL REPORTER
GAAP Targeted Improvements—Unlocking PersistencyBy Steve
Malerich
In two earlier articles (“Retrospective Noise” and “Unlocking
2.0,” The Financial Reporter, September and December 2017) I
illustrated the noise that can result from the retrospective method
when experience is consistently better or worse than assumed and I
described a technique for substantially reducing that noise. Both
articles examined effects when mortality devi-ates from the
original valuation assumption.
At the end of the December article, I noted that lapses and
sur-renders typically have a greater effect on subsequent cash
flows than on immediate cash flows. In this article, we consider
what to do when lapses and surrenders differ from expected.
LAPSE VARIANCESIn Figure 1, with early lapses much lower than
expected, earn-ings are close to ideal without any adjustment to
the reserve assumption. Since lapses align with the ultimate
assumption after a few years, there is no need for an assumption
change. If early lapses were instead higher than expected, the
requirement to write off a portion of unamortized deferred
acquisition costs (DAC) could significantly distort the earnings
pattern, but DAC amortization is outside the scope of these
articles.
[As in the earlier illustrations “Expected” shows what would
happen if experience exactly follows the original assumption,
“Ideal” shows what would happen if the original assumption had
correctly anticipated actual experience, and “Retrospective” shows
the effect of actual experience when different from the original
assumption.]
Figure 1Favorable Early Lapse Experience
-
MARCH 2018 THE FINANCIAL REPORTER | 11
In Figure 2, lapses are higher than expected by 1 percent of the
amount in force each year. Recognizing the pattern in year 8, we
unlock the assumption.
In contrast to earlier illustrations of mortality, persistent
lapse variances and the eventual assumption update have little
effect
Figure 2Persistent Adverse Lapse Experience
on net income. Even if we could extrapolate from actual
experi-ence, we wouldn’t see much benefit.
On a whole life contract, where surrenders affect cash flows
immediately and far into the future, similar experience is even
less significant to profit emergence. For the sample
Figure 3Persistent Adverse Lapse and Mortality
-
12 | MARCH 2018 THE FINANCIAL REPORTER
GAAP Targeted Improvements—Unlocking Persistency
Steve Malerich, FSA, MAAA, is a director at AIG. He can be
reached at [email protected].
whole life contract illustrated in the earlier articles, the
difference between retrospective and ideal is too small to
illustrate.
COMBINED LAPSE AND MORTALITY VARIANCESFigure 3 (page 11)
illustrates the effects of persistent adverse mortality and lapse
variances, and of unlocking both assump-tions in year six. Before
the assumption change, only mortality is extrapolated; actual
lapses are reflected as they occur.
Having seen insignificant distortions in applying the
retrospec-tive method to lapse variances, it should be no surprise
that this looks much like December’s Figure 3.
Given these illustrations, it seems likely that a formulaic
extrap-olation from actual lapse experience would cause more
problems than it would solve.
PUTTING IT INTO PRACTICEGAAP will not specify exactly when we
should update our calculations for actual experience, except that
we cannot delay beyond the annual assumption review.
Unlocking for universal life has shown us that not updating
immediately for actual experience can create confusion by
sepa-rating its effect on DAC (and SOP 03-1 reserves) from its
effect on cash flow. If anything, the problem will get worse if
applied to traditional contract reserves.
With unlocking 2.0, the reserve is less sensitive to excess
claims and there is little to gain from an immediate update for
actual claims. The net premium ratio, however, is more sensitive
and frequent updates could add volatility to new disclosures.
To realize the benefits of unlocking 2.0 without adding
vol-atility to the disclosures, I expect that many of us will find
it best to hold the net premium ratio constant in between annual
assumption reviews, updating it earlier only for espe-cially large
lapse variances. True up for actual experience, including the ratio
of accumulated excess claims to accumu-lated basis1, would be done
only during the annual assumption review process. That would
minimize disclosure volatility and have little effect on the
reserve and net income.
ENDNOTE
1 See PV (Excess Claims) in “Unlocking 2.0”, The Financial
Reporter, December 2017, page 30.
Financial Reporter N E W S L E T T E R
Help The Financial Reporter meet your needs! Please take a
four-question survey to help improve our section newsletter. Just
go to http://bit.ly/2npMkQx and give us your input!
-
SEPTEMBER 2017 THE FINANCIAL REPORTER | 13
-
14 | MARCH 2018 THE FINANCIAL REPORTER
Indexed Variable Annuities: The Next Product Frontier for the
U.S. Annuity MarketBy Simpa Baiye, Robert Humphreys and David
Knipe
Editor’s Note: This article first appeared in the February issue
of Product Matters! It is reprinted here with permission.
Indexed variable annuities (IVAs)—also known as “struc-tured” or
“buffer” annuities—are a relatively new product that have drawn
interest both among insurers and investors. IVAs have traits
insurance companies and customers find attractive, but complex
financial reporting and compliance considerations accompany them.
In order for actual and poten-tial issuers and other interested
parties to better understand the nature of these products, we
discuss in this article:
• product design,• product engineering,• issuance,• asset-
liability management, and• accounting considerations across
regulatory and GAAP
accounting frameworks.
WHAT ARE INDEXED VARIABLE ANNUITIES?Indexed variable annuities
(IVAs) (also known as “structured” or “buffer” annuities) are a
relatively new deferred annuity product. An IVA is essentially a
deferred annuity that provides equity index- linked accumulation
potential with some exposure to downside market performance. IVAs
stand in contrast to fixed indexed annuities (FIAs), which provide
limited exposure to pos-itive index returns and no exposure to
downside performance, and also to variable annuities, which provide
full exposure to market performance. Figure 1 (page 15)
demonstrates this design feature by illustrating periodic rates of
return (or cred-ited rates) for one IVA design relative to other
types of annuities and for various levels of equity market
returns.
IVA sales have grown steadily since their introduction to the
U.S. annuity market in 2012. Industry sales figures in Figure 2
(page 15) point to growing market acceptance of these
annuities.
Anecdotal surveys indicate that sales growth has been driven by
retirees and pre- retirees seeking more attractive accumulation
opportunities relative to those offered by fixed annuities and
fixed indexed annuities. We thus expect IVAs to feature more in
insurers’ product lineups in the near future.
IVA DESIGNIVAs consist of crediting accounts for renewable terms
wherein periodic interest credits (positive or negative) are linked
to the performance of a reference equity index via a formula. The
cred-iting formula places limits on upside performance that accrues
and also provides defined limits on how negative performance is
passed on to the contracts. Figure 3 (page 15) illustrates
(assum-ing that the length of the crediting strategy term is one
year) the crediting rate potential for three different crediting
designs that are prevalent as of 2017. IVA 1 provides crediting
rates that vary directly with the market and up to a predefined
limit, along with negative credits that apply to the extent that
the market drops below a defined level. IVA 2 provides crediting
rates that vary directly with market returns up to a predefined
limit with negative credits that both apply as markets drop and
level off at a defined loss level. IVA 3 provides a fixed credited
rate as long as market returns are zero or greater, along with
negative credits that apply to the extent that the market drops
below a defined level.
Early redemptions typically involve some upward or downward
adjustment to the initial deposit for the interim value of index
credits and also potentially for the market value of the bonds
backing product reserves.
Traditional variable annuity subaccounts and fixed- rate
accounts are often offered alongside IVA crediting options. In some
instances, IVAs feature limited insurance guarantees such as
guaranteed death benefits or waivers of otherwise applicable
contingent deferred sales charges.
PRODUCT ENGINEERINGThe financial building blocks for IVAs
comprise a bond com-ponent and derivatives component made up of
complementary positions in equity index options. For IVA strategy 1
illustrated in Figure 3, the IVA effectively consists of a zero-
coupon bond, a European call option that is bought, and a European
put option that is simultaneously sold. The call option provides
the upside index potential, while the put option puts the bond
investment at risk should index performance be negative. The
performance of this structure is illustrated in Figure 4 under a
variety of annual index return scenarios.
The decomposition in Figure 4 (page 16) helps clarify how
insurers could manage IVA risks. It also provides a clear path
towards interim redemption value calculations for
policyholders.
-
MARCH 2018 THE FINANCIAL REPORTER | 15
Figure 1Annuity Returns Comparison
IVA
CreditedRate
+
+
MarketReturn
–
–
VAFIA
Figure 3IVA Crediting Strategies
IVA 1IVA 2IVA 3
CreditedRate
+
+
MarketReturn
–
–
Figure 2Annuity Sales by Year
0
2
4
6
8
10
2014 2015 2016 2017 Est.
Indexed Variable Annuity Sales ($bn)
180
200
220
240
2014 2015 2016 2017 Est.
Total Annuity Sales ($bn)
65% CAGR
-4% CAGR
Source: LIMRA Secure Retirement Institute
-
16 | MARCH 2018 THE FINANCIAL REPORTER
Indexed Variable Annuities: The Next Product Frontier for the
U.S. Annuity Market
Insurer profit margins come from explicit product fees, spreads
on investments made with premium deposits, and differentials (if
any) between the revenue generated from the sale of deriv-atives
(that provide downside exposure) in excess of purchase prices of
options that provide upside market potential.
ASSET- LIABILITY MANAGEMENTBond ComponentInsurers can hedge the
bond component by investing con-tract deposits in fixed income
securities. Fixed- income investments generate yield that accrues
to the insurer and for which the insurer may take some credit,
interest- rate, and liquidity risk. The duration, liquidity and
credit risk of the bond investment should reflect product design,
the likelihood of withdrawals and redemptions, and the ongoing need
for collateral to back any derivatives traded to fund index- linked
crediting.
Derivatives ComponentInterest crediting can be hedged by
simultaneously purchasing call options with the proceeds of a
simultaneous sale of put options. The anticipated yield on fixed-
income investments may also contribute towards the purchase of call
options. Call options can be purchased on an exchange- traded or
over- the- counter (OTC) basis.
Put options can be sold on both an exchange- traded or OTC basis
to derivatives dealers. Put options could in theory also be traded
internally to meet the demand for put options to support the
hedging of existing variable annuity guarantee business.
Regulatory requirements can have a meaningful impact on the
extent to which economic asset- liability management can be
practiced. Regulation 128 in New York, as an example,
effectively places constraints on investments made with IVA
product deposits. Such regulatory limits on asset- liability risk
tolerances could indirectly influence product design options and
asset- liability management alternatives.
PRODUCT ISSUANCEThe statutory product form for an IVA would in
most cases be a modified guaranteed annuity (MGA). MGAs are
effectively deferred variable annuities which guarantee a rate of
return only if held for a defined period. Modified guaranteed
annuities are subject to regulations which impact (among other
things) product features, the creation of guaranteed separate
accounts for IVAs, and the market valuation of assets backing
reserves.
Inherent in the product design for IVAs is the possibility that
policyholders may lose part or all of their initial deposits at
con-tract maturity. For this reason, IVAs require registration
under the 1933 securities act. Issuance under securities laws is
com-plemented by the establishment of non- unitized, guaranteed
separate accounts which house assets backing reserves. These
separate accounts need to comply with relevant state laws.
Transfers between the separate account and the insurer’s general
account (as permitted) can be used to fund reserve requirements,
ongoing derivative collateral requirements, provide insurer
margins, and pay policy benefits.
US STATUTORY ACCOUNTINGThe valuation of IVA insurance
liabilities under SAP involves classifying the product within the
appropriate valuation frame-work. IVA product design and ancillary
features could be subject to valuation under Actuarial Guideline 43
(AG43) for insurance entities not effectively domiciled in New
York. However, AG43 guidelines do not provide explicit
prescriptions for the valuation
Bond Put Call Total BondValue
PutPayoff
CallPayoff
TotalValue
BondValue
PutPayoff
CallPayoff
TotalValue
BondValue
PutPayoff
CallPayoff
TotalValueValue Revenue Cost Value
Issue Date End of Year: Down Market End of Year: Flat Market End
of Year: Up Market
Figure 4IVA Building Blocks
-
MARCH 2018 THE FINANCIAL REPORTER | 17
of indexed variable annuities. As such, the specific path
towards fulfilling valuation requirements would ideally consider
both annuity minimum valuation standards and any conflicting
inter-actions with economic asset- liability management. IVAs
issued out of legal entities effectively domiciled in New York
would have reserves computed in accordance with Regulations 151
and 128.
The valuation of investments backing IVAs in the separate
account would be at market value, unless otherwise permitted by
regulators. To the extent that reserves produced by the guideline
do not share the same market sensitivity with assets backing the
same, balance sheet volatility and redundancies may occur.
US GAAP ACCOUNTINGValuation of IVA insurance liabilities under
GAAP needs to take into account the embedded derivative inherent in
the credit-ing design. As a result, ASC 815- 15, which provides
guidance on embedded derivatives, would apply and involve
identifying the host contract and embedded derivative components of
the product. The host contract would be accounted for as a debt
instrument, typically at amortized cost, while the embedded
derivative would be measured at fair value through income. An
alternative method involves valuing the entire contract (both host
contract and embedded derivative) using fair value principles by
electing the Fair Value Option based on ASC 825, financial
instruments.
Derivatives employed in hedging1 the crediting option would be
measured at fair value through the income statement. Fixed income
investments backing the IVA contract would typically be classified
as available for sale (AFS) or trading, or the fair value option
could be elected. An AFS classification for fixed income securities
involves recording unrealized gains or losses in other
comprehensive income and would be least inconsistent with a host
contract that is effectively measured at amortized cost, while a
trading securities classification or the election of the fair value
option for fixed income instruments and accounting for derivatives
at fair value would be consistent with fair valuing of the entire
annuity contract under ASC 825. A trading classifica-tion, or the
election of the fair value option for the relevant fixed income
securities would bring all realized and unrealized gains and losses
into earnings.
IMPLICATIONSIndustry sales for indexed variable annuities should
continue to grow as more insurers launch competing products in the
growing
IVA space. The design and risk- management approach for IVAs
need to balance customer needs and insurer risk appetite.
Fixed income investments and margins from the trading of
derivatives are key sources of profits for insurers. Accordingly,
the optimal investment and derivatives- use strategy for an insurer
will need to reflect product design and risk appetite, and requires
detailed analysis.
A careful analysis of accounting and valuation approaches should
occur with a clear view of the economic risk- management approach.
This analysis will serve to minimize inconsistencies between GAAP
and SAP accounting measures for both assets and IVA
liabilities.
In conclusion, IVAs represent the next potentially sizeable
opportunity for insurers to provide tax- deferred savings
oppor-tunities that meet the risk tolerances of a growing segment
of pre- retirees. We anticipate continued product innovation in
this space with the introduction of newer and more complex
crediting designs. Product transparency will need to remain
paramount as insurers manage legal and compliance risks that could
come with the proliferation of these products.
For more information, please contact the authors of this
article.
Robert Humphreys, FSA, is manager, Actuarial Services at
PricewaterhouseCoopers LLP. He can be reached at
[email protected].
David Knipe is director, Capital Markets Accounting and Advisory
Services at PricewaterhouseCoopers LLP. He can be reached at
[email protected].
ENDNOTE
1 The above does not refer to a formal designation of the hedge
relationship in accordance with ASC 815, Derivatives and
hedging.
Simpa Baiye, FSA, MAAA, is director, Actuarial Services at
PricewaterhouseCoopers LLP. He can be reached at
[email protected].
-
18 | MARCH 2018 THE FINANCIAL REPORTER
Asset Modeling Challenges for VM-20 ProjectionsBy Ben Slutsker,
Jason Kehrberg and Reanna Nicholsen
With the first year of the NAIC VM-20 transition period under
the U.S. life insurance industry’s belt, there has been significant
focus on overcoming mod-eling challenges for principle-based
reserve (PBR) valuation. In light of companies’ efforts to turn the
page from imple-menting point-in-time PBR reserves for statutory
reporting to projecting PBR reserves at future dates, this article
aims to unmask the technical challenges around asset modeling for
projecting reserves. In the following sections, we will not only
cover technical issues related to nested structures and inner and
outer loops, but also profile challenges around projected starting
assets, future hedges, negative reserves and modeling
simplifications.
FUTURE RESERVE ASSUMPTIONS: INNER VS. OUTER LOOPSA key challenge
when projecting VM-20 deterministic reserves (DR) and stochastic
reserves (SR) past the valuation date is that
we do not know what prescribed scenarios and statutory
valua-tion asset assumptions will be at future points in time.
VM-20 prescribes these assumptions for calculating reserves at the
valu-ation date, but not beyond.
Let’s first consider the situation of projecting VM-20 cash
flows for a time zero valuation. Starting Treasury rates and
spreads are based on market values observed on the valua-tion date,
and ultimate (baseline) spreads and default rates are based on
historical market averages. Starting default rates are determined
by adjusting baseline default rates for the difference between
starting and ultimate spreads, with a final adjustment if the
preliminary net spread for the entire portfolio exceeds a specified
threshold.1 VM-20 prescribes that initial spreads and default rates
grade to ultimate values by the beginning of projection year four.
Finally, future Treasury rates are generated from starting Treasury
rates using the prescribed generator.2
Now let’s consider the situation when projecting future VM-20
reserves for pricing, ALM and other internal forecasting
exer-cises. A general nested stochastic approach to project
reserves past the valuation date involves an outer loop projection
based on experience assumptions set at company discretion, and sets
of inner loop projections for each future valuation date based on
valuation assumptions. These inner loop projections not only follow
VM-20 requirements, but are also consistent with the market
environment dictated by the outer loop on the future valuation
date. In addition, each set of future inner loop projec-tions is
used to calculate a future VM-20 reserve for the outer loop
projection. This concept is illustrated in Figure 1.
Figure 1Nested Stochastic Approach for Projecting Reserves
-
MARCH 2018 THE FINANCIAL REPORTER | 19
How can we determine future valuation asset assumptions for
Treasury rates, spreads and default rates that not only follow
VM-20 requirements, but are also consistent with the market
environment dictated by the outer loop?
An inner loop projection starts with the Treasury rates and
spreads assumed by the outer loop at that point in time. The future
Treasury rate scenarios can then be generated from the prescribed
scenario generator using Treasury rates from the outer loop at that
point in time. In addition, because the ultimate spread and
baseline default rate assumptions are based on long-term
his-torical market averages, some actuaries may find it reasonable
to use the same ultimate spreads and baseline default rates that
were prescribed at time-zero for all future projected valuation
dates. Alternatively, others may prefer to modify these assumptions
to better reflect the economic conditions in the outer loop at that
time. Finally, once the baseline default rates for a future
valua-tion date are determined, the corresponding initial default
rates can be calculated using the process prescribed for the time
zero valuation.
On top of developing processes for determining future val-uation
asset assumptions, there are also challenges related to embedding
those processes within the projection model itself, which
determines the assumptions needed in VM-20 reserve projections. For
example, models may contain embedded pro-cesses for generating
future Treasury rate scenarios. But are those processes consistent
with the logic contained in the pre-scribed scenario generator? As
another example, let’s consider the process to determine starting
default rates for future inner loop projections. Depending on the
level of rigor desired, the model may need to recalculate each
asset’s weighted average life, option adjusted spread, and maximum
net spread adjustment at each future valuation date.
Finally, cash flow models at many companies make use of external
systems to project certain assets. However, there are challenges
that must be overcome when using externally pro-jected assets
(EPAs) for future inner loop projections.
• If the cash flow model relies on importing EPA files pro-duced
by the external system, the volume of data and time spent handling
it can be severe. A company can avoid this by using an application
programming interface (API) approach, which allows the modeling
platform to dynamically call the external system and read-in
external asset projections as needed.
• To preserve specific calibrations, some external systems have
limited functionality for overriding starting Treasury rates and
market values. In such cases, projecting external assets for future
inner loops may require starting at the beginning
of the outer loop, using outer loop assumptions to project to
the start of the inner loop, and using the inner loop assump-tions
thereafter.
STARTING ASSET COLLAR IMPLICATIONSVM-20 requires that the
aggregate annual statement value of starting assets, after
deducting the pre-tax interest maintenance reserve (PIMR) balance,
used to model the DR and SR must be at least 98 percent of the
final modeled reserve and no greater than the maximum of 102
percent of the final modeled reserve, net premium reserve (NPR) and
zero. Since VM-20 only applies to new business, in the early years
of PBR valuation, the level of starting assets backing the modeled
reserves may be substan-tially smaller than the actual asset
portfolio if the portfolio also supports years of business that are
outside the scope of PBR. This issue will recede over time as
pre-PBR policies terminate, but initially can have several impacts
on the projected asset portfolio used for point-in-time PBR
valuations.
A low level of starting assets due to the asset collar leads to
a larger portion of the portfolio being made up of future
pro-jected purchased assets over time. This affects the future
asset mix of the projected PBR portfolio and may cause projected
PBR portfolio rates to grade to scenario new money rates faster
than the actual portfolio would grade in reality.
This will impact both the DR and SR. Under the gross premium
valuation (GPV) method3, the DR is sensitive to the portfolio rate,
or net asset earned rate (NAER), because it is used to discount the
DR cash flows. In a low interest rate DR scenario, the projected
NAER will fall quickly, leading to a lower discount rate and a
higher DR. The DR calculated using the direct iteration method
(DIM), in addition to the SR, will also be sensitive to the
interest rate envi-ronment in each scenario, as the investment
income earned will be heavily dependent on new money rates.
Furthermore, guardrails on the modeled investment strategy, such as
requiring that fixed income reinvestment assets are no more
favorable than public non-callable corporate bonds with a credit
rating blend of 50 percent A2/A and 50 percent Aa2/AA (VM-20
Section 7.E), may drag down modeled portfolio yields when a large
portion of the portfolio is made up of newly purchased assets.
How should a company manage the discrepancy between the
projected modeled PBR portfolio and the expected actual portfolio?
To produce reasonable projections with the modeled portfolio, a
company must ensure that assumptions that rely on the portfolio
rate are aligned with the modeled portfolio rates (e.g., crediting
rates and competitor rates modeled as spreads off of the portfolio
rate).
The starting asset collar requirement creates additional
imple-mentation complexities when a company projects future PBR
-
20 | MARCH 2018 THE FINANCIAL REPORTER
Asset Modeling Challenges for VM-20 Projections
reserves. At the beginning of each inner loop projection, assets
should once again be scaled to meet the starting asset collar
requirement. Ideally, a company would project its entire block of
business, comprised of PBR and non-PBR business, up to the
projected valuation date, then scale assets to within the asset
collar before beginning the PBR projection. However,
simpli-fications may be made, such as modeling only the PBR
business in the outer loop or not rescaling at the beginning of
future inner loops.
MODELING OF DERIVATIVE PROGRAMSVM-20 requirements for modeling
derivative programs, cov-ered in VM-20 Section 7.K, are also
complex. They divide derivative programs into three types, each
with its own require-ments: clearly defined hedging strategies
(CDHS), non-CDHS hedging programs, and non-hedging derivative
programs. All existing derivative instruments already held to
support liabilities on PBR policies must be modeled, but the
treatment of future derivative instrument transactions will depend
on the type of program into which the transaction falls, which can
introduce modeling challenges.
CDHSA company is required to model future derivative
transactions associated with a CDHS. Furthermore, a company is
required to calculate an SR for any group of policies for which
there is at least one CDHS. An example of this may be an automated
hedging program for an Index Universal Life (IUL) product.
Non-CDHS Hedging ProgramsIn contrast, a company is not permitted
to model future hedging transactions that are not associated with a
CDHS. Interest-ingly, VM-20 includes a guidance note mentioning
that this requirement was added due to concerns that reserves
could
be unjustifiably reduced by including a hedging program that is
not certain to be executed. However, the guidance note also
indicates that excluding these hedging transactions may not be in
the spirit of PBR. So while VM-20 requires excluding future
non-CDHS hedging programs that decrease VM-20 reserves, it is
unclear how to treat those that increase VM-20 reserves.
Non-Hedging ProgramsFinally, a company can model non-hedging
derivative transac-tions in certain cases. If a group of policies
is excluded from the SR requirements, future non-hedging
transactions associated with those policies cannot be modeled for
the DR as per VM-20 Section 4.A.5. However, if an SR is calculated
and the derivative program is part of the company’s risk assessment
and evaluation process, future non-hedging transactions must be
modeled.
These hedging requirements, summarized in Figure 2, impact a
company’s implementation of PBR. First, a company must be prepared
to model an SR for any group of policies that employs a CDHS, even
if it would otherwise be excluded from calculat-ing an SR. As it
may take company resources to implement the SR for the first time,
valuation actuaries must know in advance if a CDHS will be added to
a group of policies. Additionally, as hedging programs that are not
part of a CDHS cannot be modeled, modeled reserves may differ from
what the actual investment strategy would indicate. However,
companies should monitor changes to VM-20 over time, as the
restrictions on modeling hedging programs could change as the
industry and regulators become more comfortable with PBR.
NEGATIVE ASSET CONSIDERATIONSAs stated, the starting assets must
be at least 98 percent of the final modeled reserve and no greater
than the maximum of 102
Figure 2 Modeling Future Derivative Programs in VM-20
The program is a non -
hedging derivative
program. Is a SR calculated for the group of policies?
Is the program part of
the company’s risk assessment and evaluation process? Must model
future transactionsProhibited from modeling future transactions in
the DRMay choose to or not to model future transactions
No YesYesNo No
-
MARCH 2018 THE FINANCIAL REPORTER | 21
Negative modeled reserves may be a common situation.
percent of the final modeled reserve, NPR and zero. So what
happens when the final modeled reserve is negative?
In the case of negative modeled reserves, a company can avoid
modeling negative starting assets by flooring at zero. However, if
the company chooses, starting assets may be negative if no less
than 98 percent of the final modeled reserve. But why would a
company choose to model negative starting assets? Intuitively,
assuming higher starting assets will generate more investment
income and lower the SR4. However, if the DR prevails, then if new
money rates are expected to increase, there may be incentive to
grade into new money more quickly at the onset of PBR
implementation. Note that negative modeled reserves may be a common
situation. Several analyses, such as the SOA VM-20 Product
Development Report5, show examples of a negative DR for term
products in early durations. Even if modeled reserves are negative,
the final PBR reserve will always be floored at the NPR, which in
turn is floored at the cost of insurance.
There are not only considerations for negative starting assets,
but also for negative future assets. Starting assets must either
cover the liquidation of benefit and expense payments (DIM) or be
set within the required range of the final reserve level (GPV
method). In both cases, there are roughly zero assets remaining
by the end of the projection. However, in theory, there may be
multiple numerical solutions to this constraint. For instance,
there may be a numerical solution in which assets become nega-tive
before returning to zero. In this situation, as invested assets
approach zero, the NAER calculated for the GPV method may
artificially inflate as the denominator decreases. This could
pro-duce an unreasonable DR level. Companies may avoid this for the
GPV method by implementing guardrails, such as ensuring the NAER is
never more negative than the borrowing rate or never more positive
than a specified yield.
POTENTIAL MODEL SIMPLIFICATION TECHNIQUESAs outlined in this
article, there are many complications for projecting future PBR
reserves. Since projected reserves are not for reporting purposes,
what simplifications can companies use? Below are common
approaches:
• Proxy Estimate: Companies can express the DR and SR as
percentage factors of the NPR or the gross premium reserve (GPR)
using best estimate assumptions. Using the NPR as a proxy may be
crude if the NPR reserve pattern varies significantly from modeled
reserves. The GPR using best estimate assumptions may serve as a
better proxy for modeled reserves, since it represents an
“un-margined” DR. If the GPR assumes best estimate assumptions,
then there is no split between inner and outer loops, making it
easier to project at future points.
• Reduced Scenarios/Policies: Use a subset of the population or,
for the SR, a subset of scenarios.
• Reduced Durations: Project reserves at periodic durations,
such as every five years, and then interpolate between. This will
reduce model run-time.
• Asset Simplifications for Non-Interest-Sensitive Business: For
products that are not sensitive to economic risk (for example,
short liability duration products such as term), assets may not
need to be modeled. Instead, a moderately adverse constant discount
rate can be assumed.
• Investment Strategy Guardrail Demonstration: For situa-tions
in which the portfolio contains a material amount of callable
bonds, the company may consider comparing the average credit
quality of a portfolio’s fixed income assets to the VM-20 guardrail
of 50 percent AA/50 percent A public
Figure 2 Modeling Future Derivative Programs in VM-20
The program is a non -
hedging derivative
program. Is a SR calculated for the group of policies?
Is the program part of
the company’s risk assessment and evaluation process? Must model
future transactionsProhibited from modeling future transactions in
the DRMay choose to or not to model future transactions
No YesYesNo No
-
22 | MARCH 2018 THE FINANCIAL REPORTER
ENDNOTES
1 The process for determining starting default rates on fixed
income assets with an NAIC designation is prescribed in VM-20
Section 9.F.1. If an asset does not have a PBR credit rating then
prescribed spreads and defaults cannot be determined and its net
yield is capped at 104% of the corresponding Treasury rate plus 25
basis points, as prescribed in VM-20 Section 9.F.5.
2 VM-20 Appendix 1 provides details on the prescribed scenario
generator, which is available in spreadsheet form on the SOA’s
website at www.soa.org/tablescalcs-tools/research-scenario/.
3 VM-20 Section 4.B (direct iteration method) describes an
approach in which com-panies may solve for starting assets that
result in the liquidation of future benefits and expenses. Section
4.A (gross premium valuation method) also provides an alternative
approach of net asset earned rate (in compliance with the starting
asset requirement in Section 7.D.2 in VM-20) to discount projected
cash flows for the reserve calculation. The two approaches should
result in solutions that are close, but may not be equal.
4 The stochastic reserve accumulates starting assets at the
projected portfolio rate and then discounts cash flows at 105% of
the 1-year treasury, per VM-20 Section 7.H.4, before subtracting
the initial starting asset amount. Therefore, the spread of the
excess of the portfolio rate over the discount rate on starting
assets results in a decrease to stochastic reserves.
5 Keating, Jacqueline. Fedchak, Paul. Rudolph, Karen. Sobel,
Uri. Steenman, Andrew. Stone, Rob. Impact of VM-20 on Life
Insurance Product Development, Society of Actuaries. Pages 20-21.
November 2016.
https://www.soa.org/Files/Research/Projects/2016-impact-vm20-life-insurance-product.pdf.
Ben Slutsker, FSA, MAAA, is a corporate vice president and
actuary at New York Life Insurance Company. He can be reached at
[email protected].
Jason Kehrberg, FSA, MAAA, is a vice president at Polysystems
Inc. He can be reached at [email protected].
Reanna Nicholsen, FSA, MAAA, is a corporate vice president and
actuary at New York Life Insurance Company. She can be reached at
[email protected].
non-callable guardrail rather than projecting reserves twice to
see which is higher.
CONCLUSIONWith VM-20 inching closer, companies should feel
encouraged to go beyond the day one big picture items, and explore
the vast terrain of nitty-gritty details required for PBR
projections. While projecting reserves at future valuation dates
may not be critical for point-in-time statutory reporting, this
capability assists companies in conducting business forecasting,
pricing and modeling economic capital in a post-PBR world.
The views reflected in this article are the views of the authors
and do not necessarily reflect the views of their employers.
Asset Modeling Challenges for VM-20 Projections
-
MARCH 2018 THE FINANCIAL REPORTER | 23
Leveraging X-factor Testing Techniques in Developing Mortality
Assumptions for VM-20By Jerey R. Lortie and Ying Zhao
VM-20 requires prudent estimate assumptions that are based upon
a combination of company experience, industry basic tables and
prescribed margins. VM-20 has requirements on how company
experience is defined, and also requires that additional margins be
established if the actuary does not consider the prescribed margins
to be adequate. The use of company-based assumptions for statutory
valuation and performance of certain tests on that mortality have
been in place since Regulation XXX became effective, and some of
the tools used within X-factor testing can be leveraged for use
with VM-20. In this article, we will connect VM-20 to Regula-tion
XXX and repurpose some of the techniques that actuaries have been
using for X-factor testing to aid in setting VM-20 mortality
assumptions.
REVIEW OF VM-20 AND REGULATION XXXRequirements for
Principle-Based Reserves for Life Products (VM-20) took effect for
direct writers on Jan. 1, 2017, with the adoption of the Valuation
Manual. VM-20 contains a three-year transition period so that by
Jan. 1, 2020, all newly issued policies must be valued in
accordance with VM-20. Generally speaking, VM-20 implementation is
(or will be) a long and strenuous process, requiring a great deal
of rigor, judgment and documentation. One of the many challenges
companies face is the development of prudent estimate mortality
assumptions as required in Section 9.C. In addition to the
mechanical require-ments, VM-31 (which includes requirements for
disclosures of assumptions within the PBR Actuarial Report) and
VM-G (which covers corporate governance regarding principle-based
reserves) bring assumption-setting for booked statutory reserves
under greater scrutiny.
Assuming mortality segments have been defined, the prudent
estimate mortality assumption process can be summarized in a few
steps, as is done in Section 9.C.1:
1. Develop company experience mortality rates,
2. determine industry basic table to which company experience
mortality rates will grade,
3. determine credibility of underlying company experience,
4. determine prescribed margins and
5. blend company experience mortality rates and industry basic
table according to grading period determined.
While several aspects of the prudent estimate mortality
assump-tion are prescribed (selection of industry basic table,
margins, grading), the process of determining company experience
mor-tality rates according to Section 9.C.2 is less rigidly
defined. However, there are several governing requirements outlined
within VM-20:
• Annual Assumption Review and Validation via Statistical
Testing
From 9.A: “The company shall use its own experience, if relevant
and credible, to establish an anticipated experience assumption for
any risk factor…
The appointed actuary shall annually review relevant emerging
experience for the purpose of assessing the appro-priateness of the
anticipated experience assumption. If the results of statistical or
other testing indicate that previously anticipated experience for a
given factor is inadequate, then the appointed actuary shall set a
new, adequate, anticipated experience assumption for the
factor.”
• Company Experience Mortality Rates used in VM-20 are at or
greater than best estimate
From 9.C.2.c: “The company experience mortality rates shall not
be lower than the mortality rates the company expects to emerge
which the company can justify and which are disclosed in the PBR
Actuarial Report.”
• Further analysis required to determine if the prudent
esti-mate mortality assumption is sufficient
From 9.C.5.d: “The prescribed margin percentages shall be
increased, as appropriate, to reflect the level of uncertainty
related to situations …” ASOP 52, Section 3.4.6, further
clarifies:
“a. Mortality Margins—Section 9 of VM-20 prescribes
the margins that are to be added to the anticipated
experience mortality assumptions but also requires the
establishment of an additional margin if the
prescribed margin is inadequate. The actuary should use
professional judgment in determin-ing such additional margin.
The guidance in the remainder of this section on determining
assumption margins does not
-
24 | MARCH 2018 THE FINANCIAL REPORTER
Leveraging X-factor Testing Techniques in Developing Mortality
Assumptions for VM-20
apply to the prescribed mortality assumptions but does apply
when determining additional margins for mortality.
“b. Establishing Margins—For each assumption that includes
a margin, the actuary should reflect the degree of risk and
uncertainty in that assumption in determining the magnitude of
such margin. When determining the degree of risk and
uncertainty, the actuary should take into account the magnitude and
frequency of fluctuations in relevant expe-rience, if
available. In doing so, the actuary should consider using
statistical methods to assess the potential volatility of the
assumption in setting an appropriate margin.”
There are direct connections between the wording of VM-20 and
Regulation XXX regarding the assessment of the appropri-ateness of
the company experience assumptions. Section 5B(3)(g)(iii) of
Regulation XXX calls for review of continued appro-priateness of
X-factors taking into account relevant emerging experience, a
method generally known as “retrospective testing,” which aligns to
the wording in Section 9.A of VM-20. Section 5B(3)(c) of Regulation
XXX provides the requirement that mortality rates over a period of
time from the valuation date must exceed company best-estimate
(which, when combined with other requirements spelled out in
Regulation XXX, comprise “prospective testing”), comparable to the
wording in
Section 9.C.2.c of VM-20. The retrospective testing might also
help the actuary understand the volatility around the company
experience mortality rates, which is helpful in understanding the
appropriateness of prescribed margins outlined in Section 3.4.6(a)
of ASOP 52.
REVISITING X-FACTOR TESTING TECHNIQUESIn the context of
Regulation XXX, retrospective testing provides insight as to
whether or not emerging experience supports the use of a particular
set of X-factors. Typically, the test involves building a
statistical distribution of claims based upon the X-factors being
tested and determining whether actual claim experience is an
outlier in that distribution, generally at or above the 95th
percentile. Similarly, building a statistical distribution of
claims may be useful for VM-20. By generating a claim dis-tribution
where the expectation is based upon a proposed set of company
experience mortality rates, the actuary can benchmark where actual
experience lies on the distribution to assess appro-priateness of
the proposed rates. In addition, the actuary may assess volatility,
distinguish fluctuation from a change in trend of emerging
experience, or identify the percentile ranking of claims emerging
according to the prudent estimate mortality to determine whether
additional margins are needed according to Section 9.C.5.d of
VM-20.
Two tools, prevalent in X-factor testing, are used to build out
the claim distribution. The Panjer recursive method is an algorithm
designed to build a distribution based upon grouped data (e.g.,
face amount bands), and has the advantages of being formulaic and
repeatable (which auditors and reviewers appreciate). The
distribution is initialized with the probability of zero claims and
builds from there. The reader is directed to a pair of write-ups
that are of great value: the original article1 by Harry H. Panjer
which develops the method, and a later article2 by Lloyd Spen-cer
which provides an excellent illustrative example.
The other tool, Monte Carlo simulation, is based on randomly
generated numbers and can better emulate the true distribution if
given enough trials and seriatim data. For each policy, within a
single trial, a random number is drawn between 0 and 1. If the
random number is less than the mortality rate for that policy, then
a death is assumed to occur, and the sum of the deaths across all
policies provides the claims for a single trial. Then, the process
is repeated for a particular number of trials, usually a number
large enough so that the randomness of the number generation does
not materially alter the result (typically 10,000). Results are
then ordered and the distribution created. While the Monte Carlo
method is possibly a better representation of the true claim
dis-tribution, at least if done at a seriatim level to capture
individual policy expected mortality, it is calculation-intensive
and more challenging to audit due to its random number
generation.
-
MARCH 2018 THE FINANCIAL REPORTER | 25
Within the Regulation XXX framework, prospective testing is a
two-step test to determine if X-factor mortality is at least as
great as best estimate. Starting with the in force policies subject
to XXX as of a particular date, the following calculations are
performed, using both X-factor and best-estimate mortality (without
mortality improvement beyond the valuation date):
• Calculate the actuarial present value of future death benefits
to the end of the first segment per policy and
• calculate mortality rates without recognition of mortality
improvement beyond the valuation date, in each of the first five
years after the valuation date.
In both steps, the metric computed using X-factor mortal-ity
must exceed the same metric based upon best-estimate. Translating
to VM-20, the actuary could perform similar tests, projecting out
to the point where the prudent estimate is fully dependent upon the
industry-basic table (i.e., after the grading has completed), which
could cover the requirement of Section 9.C.2.c of VM-20 by
comparing the mortality rates based upon the company experience
mortality rates to best-estimates to ensure that the former rates
exceed the latter. The actuary could also use this technique to
confirm that the prudent estimate mortality rates (post-grading)
exceed best-estimate, especially in later projection years, where
old-age mortality assumptions are commonly graded to industry
averages.
CONCLUSIONEstablishing prudent estimate mortality assumptions
under VM-20 is a long and complex process. Having the
ability to generate company experience mortality rates, creating
a mechanism to determine the validity of these assumptions, and
developing a manner in which to benchmark the prescribed mortality
margins for adequacy, will be critical components of the VM-20
process. While still a significant undertaking, techniques from
Regulation XXX can be repurposed to address these chal-lenges.
ENDNOTES
1 “The Aggregate Claims Distribution and Stop-Loss Reinsurance”
published in the Transactions of the Society of Actuaries, Volume
XXXII, 1980, pages
523–545.)https://www.soa.org/library/research/transactions-of-society-of-actuaries/1980/january/tsa80v3215.pdf
2 “An Overview of the Panjer Method for Deriving the Aggregate
Claims Distribution,” Lloyd Spencer
http://www.actuary.org/pdf/external/panjer_spencer.pdf
Je® rey R. Lortie, FSA, MAAA, is AVP –Valuation and Financial
Reporting for Lincoln Benefit Life in Rosemont, Ill. He can be
contacted at je [email protected].
Ying Zhao, FSA, MAAA, is AVP –Product Management for Lincoln
Benefit Life in Rosemont, Ill. She can be contacted at
[email protected].
-
26 | MARCH 2018 THE FINANCIAL REPORTER
IFRS 17 Variable Fee ApproachBy Tze Ping Chng, Steve Cheung and
Anson Yu
After a very long journey, the International Accounting
Standards Board (IASB) issued IFRS 17 Insurance Contracts (IFRS
17). IFRS 17 replaces IFRS 4 that was issued in 2004. The overall
objective is to provide a more use-ful and consistent accounting
model for insurance contracts among entities issuing insurance
contracts globally.
GENERAL MODEL AND VARIABLE FEE APPROACHThe IASB introduces a
general accounting model (GM, previ-ously called
building-block-approach) for the insurance contract liability
measurement.1 In order to cater to the unique features of insurance
contracts with direct participation features, IFRS 17 provides for
a specific approach called the variable fee approach (VFA).
Insurance contracts with direct participation features (or “direct
participating contracts”) are insurance contracts that are
substantially investment-related service contracts under which an
entity promises an investment return based on underlying
items. These may be regarded as creating an obligation to pay
policyholders an amount that is equal to the fair value of the
underlying items, less a variable fee for service.
VFA is a modification of GM in order to reflect the nature and
economics of these direct participating contracts. Table 1
sum-marizes the key differences between GM and VFA.
VFA ELIGIBILITY CRITERIAThe IASB made it clear that only
insurance contracts with direct participation features are eligible
for the VFA, but significant judgment is required to assess the VFA
eligibility, as outlined in paragraph B101 and BC238.
Insurance contracts with direct participation features are
insur-ance contracts for which, on inception:
a. the contractual terms specify that the policyholder
partic-ipates in a share of a clearly identified pool of underlying
items; (VFA criteria I)
b. the entity expects to pay to the policyholder an amount equal
to a substantial share of the fair value returns from the
underlying items; (VFA criteria II) and
c. the entity expects a substantial proportion of any change in
the amounts to be paid to the policyholder to vary with the change
in fair value of the underlying items. (VFA criteria III)
Measurement model
Changes in fulfilment cash flows (FCF) due
to the changes in financial variables
Insurance finance income or expenses
GM All changes in dis-count rates and other financial variables
are reported in the statement of compre-hensive income
The interest expenses on the contractual service margin (CSM)
are explicitly accreted using rates at the initial recognition of
the contracts
VFA CSM is adjusted to reflect the changes in the variable fee,
which includes some changes in discount rates and other financial
variables
The interest expenses are implicit in the changes in the
insurer’s variable fee
Table 1Differences between GM and VFA
-
MARCH 2018 THE FINANCIAL REPORTER | 27
The IASB made it clear that only insurance contracts with direct
participation features are eligible for the variable fee approach,
but significant judgment is required to assess the VFA
eligibility.
KEY CONSIDERATIONS FOR LIFE INSURERS With the significant
judgment required for the VFA eligibility assessment, we expect
these to be part of the key issues to be discussed by the
Transition Resource Group (TRG). Below are some evolving questions
we observed from the market.
1. What is a clearly identified pool of underlying items?
The pool of underlying items can comprise any items, for
exam-ple a reference portfolio of assets, the net assets of the
entity, or a specified subset of the net assets of the entity, as
long as they are clearly identified by the contract. An entity need
not hold the identified pool of underlying items because the
mea-surement of insurance contracts should not depend on what
assets the entity holds. The underlying items do not need to be a
portfolio of financial assets. They can comprise items such as the
net assets of the entity or a subsidiary within the group that is
the reporting entity.
2. What is the definition of “contract” and “contractual terms”
when defining the clearly identified pool of assets?
A contract is an agreement between two or more parties that
creates enforceable rights and obligations. Enforceability of the
rights and obligations in a contract is a matter of law. Contracts
can be written, oral or implied by an entity’s customary business
practices. Contractual terms include all terms in a contract,
explicit or implied. Implied terms in a contract include those
imposed by law or regulation.
There are certain features which may not satisfy VFA criteria I:
(i) different portfolios of participating contracts (direct or
indirect) share the same fund with notionally separated assets in
the entity’s general account, and (ii) the segregation of assets
are only managed internally without enforceability or proper
disclosure to the policyholders. While “ring-fenced-asset” may
better meet this criterion, there are also discussions if the
“accounting designation” or “entity’s governance framework and
disclosure” meet this criterion. Advocates argue that com-mercial
communication, i.e., materials presented or disclosed to the
policyholders, can form part of the enforceability and the entity
should consider these factors for the assessment of clearly
identified pool of assets. In any case, the definition of the
“underlying items” should be documented clearly, and the entity
cannot change the underlying items with retrospective effects.
3. Does “a share of a clearly identified pool of underlying
items” preclude the entity’s discretion to vary the amounts paid to
the policyholder?
No, but the link to the underlying items must be
enforceable.
4. How to interpret the word “substantial” in VFA criteria II
and III?
The IASB does not provide a concrete definition for the term
“substantial” as noted in the VFA criteria II and III. This is to
allow entities to apply IFRS 17 for their particular circum-stances
without being limited by any quantitative rules. We expect that
market consensus will converge with potential help from TRG
discussion. However, a range of sharing percentages may still be
expected from various jurisdictions due to different product
offerings, and comparability with the fee structures of the
investment products offered. An individual entity needs to perform
its own assessment, and verify its conclusion with its respective
auditor.
5. What is the “variable fee”?
A variable fee that the entity will deduct in exchange for the
future service provided by the insurance contract, comprises: (i)
the entity’s share of the fair value of the underlying items; less
(ii) fulfilment cash flows that do not vary based on the returns on
underlying items. Contracts eligible for VFA should specify a
determinable fee which can be expressed as a percentage of
portfolio returns or asset values rather than only as a monetary
amount. Without a determinable fee, the share of returns on the
underlying items the entity retains would be entirely at the
discretion of the entity, and this would not be consistent with
that amount being equivalent to a fee.
ILLUSTRATIVE EXAMPLEA simple five-year investment-linked product
is created to illustrate the CSM differences between VFA and GM,
with the projection given in Table 2:
• Death benefit (sum assured) = fixed 500 + account value (AV),•
maturity benefit = AV,• level annual premium = 500,• 2 percent
asset management charge (AMC),• cost of insurance charge (COI
charge) and• 100 identical policies issued.
-
28 | MARCH 2018 THE FINANCIAL REPORTER
IFRS 17 Variable Fee Approach
Projected best estimate cash flows (BECFs) for 100 policies at
inception
BE projection/Year Yr1 Yr2 Yr3 Yr4 Yr5 Yr6
No. survival(BOY) 100 99 97 94 90 85
No. deaths(EOY) 1 2 3 4 5
No. survival(EOY) 99 97 94 90 85
Premium(BOY) 50,000 49,500 48,500 47,000 45,000 -
Commission(BOY) 5,000 2,475 970 - - -
Expense(BOY) 200 198 194 188 180 -
Death outgo(EOY) 1,000 3,030 6,186 10,638 16,667 -
Survival outgo(BOY) - - - - - 240,833
Net CF 43,800 43,797 41,150 36,174 28,153 (240,833)
Projected policyholder AV (PHAV) for 100 policies at
inception
PHAV(BOY) - 49,500 98,470 146,814 194,362
Premium(BOY) 50,000 49,500 48,500 47,000 45,000
COI charge(BOY) (500) (1,000) (1,500) (2,000) (2,500)
Investment income(EOY) 1,490 4,460 8,940 15,022 22,876
AMC(EOY) (990) (1,960) (2,909) (3,836) (4,737)
Death outgo from PHAV(EOY) (500) (2,030) (4,686) (8,638)
(14,167)
PHAV(EOY) 49,500 98,470 146,814 194,362 240,833
Table 2Assumed projected cash flows and AV
-
MARCH 2018 THE FINANCIAL REPORTER | 29
Table 3 summarizes the key steps in calculating the initial and
subsequent CSM under GM as follows:
1. FCF and initial CSM are the same under both GM and VFA. In
this example, the cash flows for the FCF calculation are based on
the BECFs with the following assumptions: (i) directly attributable
expenses = 100% BE expenses, and (ii) investment component = death
and survival outgo supported by PHAV.
2. FCF is the PV of the risk adjusted cash flows which includes
best estimate liability (BEL) and risk adjustment (RA). The
discount rate (initial DR) is assumed to be the PHAV growth rate
(which in this example is the risk-free yield curve).
Initial CSM is the unearned profit at inception and is equal to
the negative of FCF floored by zero. For simplicity, RA is assumed
to be zero.
3. The number of coverage units in a group is the quantity of
coverage provided by the contracts in the group, determined by
considering for each contract the quantity of the benefits provided
under a contract and its expected coverage dura-tion. In this
example, it is assumed to be the number of policy in force * sum
assured (including AV).
4. The BOY CSM is accreted with interest (at initial DR), and
then amortized according to the coverage unit pattern.